True UX professionals use different types of persona for different contexts: the assumption persona (as you describe) is a quick shorthand list of assumptions, and completely worthless unless used as a basis for designing experiments (e.g. data analysis) to test those assumptions.

I'm a big fan of qualitative personas: you create your assumption persona and go as wild as you like with the vivid descriptions. Then you conduct depth interviews - at least an hour - with at least 5 examples of each persona type, so you'd speak to 6 or more Grinders, 6 or more Experts, etc. and ideally - if you can do so - repeat the exercise across different countries and contexts.

Then - where the proverbial magic happens - you go through your assumption personas line by line and compare them to your interview findings, and note all the things you got wrong. If you did it well, it will be a long list!

Statistical personas, as you describe, are also an effective way of segmenting your audience, but (as you say) it requires a lot of data and it can be quite expensive if you need to run surveys with thousands of participants, which you'd have to do if you don't have access to actual user data.

In most cases, qualitative (interview-based) personas strike a useful balance between imagination and hard data.